Recently, Beijing News·Beike Finance published a feature titled “27 Innovation Consortia Established: In Beijing, They Are Finding Partners for Technical Breakthroughs.” The report revealed major progress from the “3C Smart Manufacturing Innovation Consortium” led by Xiaomi Group. Within this initiative, YMatrix (Beijing Siwei Zongheng), a domestic database vendor, collaborated with Xiaomi for two years to solve a highly demanding industrial data challenge: real-time ingestion of 1.7 billion records per day and sub-second analytics across hundreds of production dashboards in Xiaomi’s smartphone manufacturing plants.
The joint breakthrough now stands as a replicable blueprint for building a domestically developed, industrial-grade data foundation.
In Xiaomi’s automated smartphone factories, the production lines operate in a mode defined by high concurrency, multi-point data collection, and 24/7 non-stop throughput. This environment generates a continuous data deluge:
• More than 1,000 machines across the factory • Each machine equipped with thousands of sensors • Pressure, temperature, current and other time-series signals uploaded every second • Tens of thousands of records produced each second • A total of 1.7 billion records written every day
These massive data streams must not only be ingested in real time but also support sub-second queries across hundreds of dashboards for production, quality, and equipment monitoring. Legacy systems once suffered from write bottlenecks and slow queries, even triggering production risks. Xiaomi urgently needed a domestically developed, high-resilience data foundation capable of withstanding this extreme data pressure.
In 2022, Xiaomi initiated the “3C Smart Manufacturing Innovation Consortium,” bringing together more than 40 universities, research institutes, and industry partners including Tsinghua University and the Chinese Academy of Sciences. The consortium decomposed complex manufacturing challenges into solvable technical topics and sought the right partners for each. YMatrix was selected to lead the core track: building the industrial data foundation.
Working against real production data from Xiaomi’s lines, YMatrix and Xiaomi teams ran multiple rounds of extreme stress testing and iterative optimization.
Key enhancements included: • MatrixGate for high-throughput time-series ingestion directly into DWD detail layers, allowing the system to absorb peak concurrent writes. • MARS3 hybrid storage engine for optimized time-series compression and write latency stabilization within tens of milliseconds. • Unified time-series analytics architecture, replacing the traditional multi-engine pipeline (time-series DB + OLAP engine + wide-table processing) with in-database analytics to eliminate cross-engine performance loss.
After nearly two years of joint development, the system not only handles Xiaomi’s daily 1.7 billion data writes but also delivers sub-second dashboard queries, achieving over 99.9% stability in continuous production.
Xiaomi’s Director of Industrial Big Data, Liang Yaoting, commented: “Together, we resolved the challenges of massive data ingestion and real-time analytics in Xiaomi’s factories.”
YMatrix Founder and CEO, Yandong Yao, noted: “Working side-by-side with Xiaomi allowed us to refine YMatrix directly in real production environments, significantly enhancing stability and performance for smart manufacturing scenarios.”
YMatrix defines itself as a Hyper-Converged Database, built on a multi-engine microkernel architecture that supports relational, time-series, JSON, GIS, and other data models within a single system. It provides unified modeling and analytics through standard SQL and enabled Xiaomi to establish three core capabilities:
MatrixGate is optimized for high-frequency, high-volume time-series writes. MARS3 provides superior compression and write performance, ensuring zero backlog and no write pressure buildup.
YMatrix offers much richer analytics than traditional time-series databases, covering line monitoring, alarms, BI dashboards, visualization walls, and more. It also supports in-database machine learning through PL/Python, enabling applications such as predictive maintenance.
One database replaces the typical combination of online DB + time-series DB + analytical DB. This significantly reduces architectural complexity, maintenance overhead, and infrastructure costs.
The Xiaomi collaboration demonstrated not only a robust industrial data foundation but also validated YMatrix’s capabilities in real production environments. The resulting trio of strengths—reliable time-series ingestion, flexible analytics, and simplified architecture—are now being replicated across industries such as energy & utilities, smart EV manufacturing, and industrial IoT. This provides a practical, scalable adoption model for domestic databases in intelligent manufacturing.
As a flagship example of Beijing’s innovation consortium model, the Xiaomi 3C Smart Manufacturing Consortium embodies “enterprises posing problems, partners solving them,” accelerating deep industry-academia-research collaboration. YMatrix’s contribution marks an important milestone in the scaled application of domestic databases in industrial environments.
From lab to production line, from technical breakthroughs to ecosystem collaboration, more solution providers like YMatrix are becoming key drivers of China’s intelligent manufacturing upgrade.
Media & Partnership Contact: YMatrix Marketing & Ecosystem: [email protected]
Beijing Siwei Zongheng Data Technology Co., Ltd. (YMatrix) was founded in August 2020. Its core team joins from leading companies such as Oracle and Greenplum, with several members being key contributors to the Greenplum kernel.
YMatrix focuses on next-generation data infrastructure for the IoT era and is dedicated to building a Hyper-Converged Database that supports unified TP+AP+AI workloads and large-scale enterprise data warehouses.
YMatrix currently serves customers including CATL, BYD, Li Auto, Xiaomi, ZTE, and more, with benchmark deployments in finance, manufacturing, energy, and automotive industries.
How YMatrix Domino Replaces Lambda, Kappa, Flink, and Spark with One Engine🚀
Ganfeng Lithium: Real-Time Reporting at Scale with YMatrix
How a Leading ERP Vendor Entered the AI Fast Lane — A YMatrix Field Story
Dahshenlin: Achieving Real-Time Finance-Operations Integration with a Modernized Data Foundation
SERES × YMatrix: 3-Hour Migration of 2.13TB, 50% Faster Multi-Scenario Queries
